Early-career researcher spanning multimodal game-agent benchmarks and generative video world models
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-08) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored: publication counts, venues, and output trajectory are unavailable in the current bibliometric snapshot. This axis would measure total and per-year research production; the only dated evidence now is two 2026 arXiv papers.
Impact—
Not yet scored: citation counts, h-index, and uptake signals are absent from the current data. This axis would measure how widely the two catalogued papers are cited and built upon by other researchers.
Focus0/100
Scores 0, reflecting a perfectly even split: the field distribution is 50% machine learning (multimodal game agents and benchmarks) and 50% computer vision (generative video world models). The two catalogued 2026 papers sit one in each field, so the record shows breadth across two adjacent generative-AI areas rather than concentration in a single line of work.
Collaboration62/100
Scores 62, based on 22 distinct coauthors in the record, 12 of whom are flagged as frequent collaborators — including Wangbo Yu, Ying Shan, Xingyilang Yin, Junfu Pu, and Xiaodong Cun. This points to stable, team-based work on generative-AI projects, though the current data does not map individual coauthors to specific papers.
Activity—
Not yet scored: per-year publication counts and recency signals are missing from the current snapshot. This axis would measure publishing cadence and ongoing research tempo; at present the record shows two arXiv releases both dated 2026.
Reproducibility38/100
Scores 38, the mean readiness across 2 reviewed papers. GameHorizon Suite rates medium: a large benchmark release supports reuse, though its VLM-generated ground truth needs scrutiny. WorldCrafter rates low: the headline result is backed mainly by a self-built benchmark. Reproducibility thus varies noticeably between the two outputs.
Research summary
Yiran Wang is an early-stage researcher whose catalogued output spans two adjacent generative-AI areas: multimodal game agents and benchmarks within machine learning, and generative video world models within computer vision, split evenly 50/50 in the field distribution. The Evolve catalog contains two 2026 arXiv papers. The first, GameHorizon Suite, has Wang as first author: a multi-horizon gameplay dataset and evaluation benchmark reported to cover 5k hours of play across 21 games with 47 models evaluated. Evolve's verdict credits the scale but flags two scrutiny points — VLM-generated ground truth and the absence of online AAA-game evaluation. The second, WorldCrafter, places Wang in a middle author position: a video world model with implicit 3D-aware memory for long-horizon consistency, reporting a 47.6% revisit-consistency gain over 8 baselines with strong ablations; Evolve notes the result rests mainly on a self-built benchmark and rates reproducibility readiness low. Radar scores: collaboration 62 (22 distinct coauthors, 12 recurring, including Wangbo Yu, Ying Shan, and Xiaodong Cun), reproducibility 38 (mean across 2 reviewed papers), and focus 0, reflecting the even two-field split. Output, impact, and activity lack sufficient data and remain unscored. Overall this is a full but early profile, with evidence concentrated in 2026 preprints and a two-track agenda pairing benchmark building with world-model architecture.
Collaboration pattern
The record shows 22 distinct coauthors, with 12 recurring — including Wangbo Yu, Ying Shan, Xingyilang Yin, Junfu Pu, and Xiaodong Cun — a pattern consistent with team-based generative-AI research rather than isolated work. The collaboration score of 62 reflects moderately broad but recurring partnerships; the current data does not link specific coauthors to individual papers, so internal role structures remain unclear.
Research keywords
Highlights
First-authored GameHorizon Suite: 5k-hour AAA gameplay benchmark
WorldCrafter: 3D-aware memory lifts video world-model consistency
Timeline
- 2026First-authored GameHorizon Suite on arXiv — a multi-horizon gameplay dataset and evaluation benchmark spanning 5k hours across 21 games with 47 models evaluated.
- 2026Co-authored WorldCrafter on arXiv — a consistent video world model with implicit 3D-aware memory, reporting a 47.6% revisit-consistency gain over 8 baselines.
- 2026Both catalogued outputs land in 2026 across two distinct fields — machine learning game-agent benchmarks and computer vision world models — indicating a parallel two-track agenda.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Radar scores are computed from public bibliometric data only; output, impact, and activity axes lack sufficient data and are unscored.
- 'Yiran Wang' is a common name; this profile is name-resolved from public bibliographic data and may merge or split same-name authors.
- Both catalogued papers are 2026 arXiv preprints; peer-review status, institutional affiliation, and career details are not present in the current data.
- Reproducibility score and verdicts reflect Evolve's review of only 2 papers (mean 38), not independent replication results.